Forcibly driven coherent soft phonons in GeTe with intense THz-rate pump fields

Forcibly driven coherent soft phonons in GeTe with intense THz-rate pump fields

Muneaki Hase and Masahiro Kitajima
Materials Engineering Laboratory, National Institute for Materials Science, 1-2-1 Sengen, Tsukuba 305-0047, Japan

o

(N

C/3

Shin-ichi Nakashima
Power Electronic Research Center, National Institute of Advanced Industrial
Science and Technology, 1-1-1 Higashi, Tsukuba 305-8561, Japan

Kohji Mizoguchi
Department of Applied Physics, Osaka City University,
3-3-138 Sugimoto, Sumiyoshi-ku, Osaka 558-8585, Japan

We propose an experimental technique to generate large amplitude coherent phonons with irradi-
ation of THz-rate pump pulses and to study the dynamics of phase transition in GeTe ferroelectrics.
When a single pump pulse irradiates the sample at various pump power densities, the frequency of
the soft phonon decreases sub-linearly and saturates at higher pump powers. By contrast, when
THz-rate pump pulse sequence irradiates the sample at matched time intervals to forcibly drive the
oscillation, a large red-shift of the phonon frequency is observed without saturation effects. After
excitation with a four pump pulse sequence, the coherent soft phonon becomes strongly damped
leading to a near critical damping condition. This condition indicates that the lattice is driven to a
precursor state of the phase transition.

PACS numbers: 78.47.-(-p, 77.84.-s, 42. 65. Re, 63.20.-e,

o

>

m

X

Femtosecond-attosecond laser technologies have re-
cently been the focus of much attention in solid state
physics because of their growing applications in observ-
ing both electronic and phononic ultrafast dynamics. In
particular, one of the possibilities for the application of
femtosecond laser pulses is controlling the amplitudes of
coherent collective motions of atoms excited in condensed
matter byusing double pump pulses or multiple pump
pulses. [T-4] One can enhance the amplitude of a phonon
mode by applying in — phase pulses, or suppress the am-
plitude by applying out — of ~ phase pulses, both phe-
nomena are observed in a real time domain. The most
remarkable advantage of the multiple pulse pump tech-
nique is the ability to avoid saturation effects due to high
density excitation , i.e., screening of the space-charge field
by electron-hole plasma. Q

One goal in controlling coherent lattice vibrations is
to cause lattice instabilities which could lead to a phase
transition. We note that manipulation of phonons cannot
be demonstrated with conventional frequency-domain
spectroscopy. Until now, all experiments in coherent con-
trol of phonons have been performed under low density
excitation, in which the coherent phonon amplitude was
too small to observe lattice instability. [ll-Q In this letter,
we propose an experimental technique to forcibly drive
multiple coherent phonons into one larger amplitude co-
herent phonon and demonstrate its capability of inducing
an extremely unstable crystal phase close to the critical
point in ferroelectric materials. We used a twin Michel-
son interferometer to produce intense femtosecond THz-
rate pulse trains, which were used to repetitively push
lattice motions in-phase. Thus, a larger amplitude coher-

ent phonon was generated without any saturation effects.
The coherent phonon with larger amplitude showed a
large decrease in frequency and became strongly damped,
showing that the lattice is driven into a strongly anhar-
monic regime near the structural phase transition.

We chose ferroelectric GeTe as a sample because of
its considerable interest in optical memories applications
due to its reversible structural changeQ: This crystal un-
dergoes the rhombohedral-rocksalt structural change at
the transition temperature Tc = 657±100 K, attributed
to a displacive phase transition. Q In a displacive phase
transition, the motions between the two phases gener-
ally involve a soft mode vibration, whose frequency is
dramatically reduced near Tc.[8] This type of the phase
transition is characterized by a single potential minimum,
whose position shifts at Tc. By contrast, other type of the
phase transition, such as a order-disorder transition, is
characterized by several potential minima among which a
choice is made at Tc.fsj The order-disorder transition oc-
curs with collective tunneling or thermally assisted hop-
ping modes. In the case of GcTc, the Ai mode has been
considered to be the soft mode, which was observed by
Raman measurements. 7] An ab initio theoretical inves-
tigation predicted that the phase transition in GeTe was
a fluctuation-driven first-order phase transition. [9]

Because of the strongly reduced frequency of the soft
mode near Tc, monitoring the phase transition dy-
namics is difficult with conventional frequency-domain
spectroscopies. 10] Motivated by these difficulties. Nel-
son and coworkers examined time-resolved pump-probe
measurements at various lattice temperatures in per-
ovskites, and the heavily damped soft phonon was ob-

2

i

<

12.7 mJ/cm

o data
— fit

0.0 1.0 2.0

Delay time (ps)

3.0

FIG. 1: The transient reflectivity change obtained at 12.7
mJ/cm^. The open circles are the experimental data, the
solid curve is the fit to the data with Eq. (1). The inset
shows the amplitude of the coherent oscillation obtained by
the fitting.

served near T^.[ll| The present study approaches the
critical point by increasing phonon amplitude instead of
increasing lattice temperature. The sample studied in
this work was a single crystal of GeTe prepared by a va-
por growth method and cleaved in the c crystallographic
plane. GeTe is a narrow band-gap semiconductor, and
the generation of the coherent Ai phonon is closely re-
lated to excitation of carriers from the valence band to
higher energy bands, i. e., displacive excitation of coher-

ent phonons (DECPs). IJjllSl Femtosecond time-domain
measurements were carried out at room temperature with
a pump-probe technique. Femtosecond laser pulses of a
Ti:sapphire laser, operating at 800 nm, were amplified
to a pulse energy of 500 /iJ in a 1 kHz regenerative am-
plifier. After compensation of the amplifier dispersion,
the amplified pulses had 120 fs duration. The pump and
probe beams were focused on samples to a diameter of
about 100 /im. The pump power density was reduced by
a neutral density filter to below 13 mJ/cm^ to prevent
damaging the sample or causing laser ablation, and it was
varied from 0.8 to 12.7 mJ/cm^. The probe pulse energy
was also reduced and fixed at 0.3 mJ/cm^. The pump-
beam was mechanically chopped at 315 Hz for the signal
detection by a lock-in amplifier. The transient reflectiv-
ity change AR/R was recorded by changing the optical
path length of the probe beam.

Figure 1 shows time derivatives of the transient re-

flectivity change obtained by a single pump pulse at the
pump power density of 12.7 mJ/cm^. The coherent os-
cillations due to the collective motion of the crystal lat-
tice appear on the slowly varying background due to the
photo-excited carriers. The frequency and the amplitude
were obtained by fitting the time domain data using a
damped harmonic oscillator with the background.

F{t) = Ae"*/^cos(a;i + 0) + B[e

(1)

where A and B are the amplitude of the coherent phonon
and the carrier contributions, respectively, r is the de-
phasing time of the coherent phonon, ri and T2 are the
relaxation and rising times of the carrier background, re-
spectively. Lo is the frequency and (p is the initial phase
of the coherent oscillation. The time period of the oscil-
lation at the lowest pump power density of 0.8 mJ/cm^
is ~ 263 fs (= 3.80 THz), which is close to that of the Ai
mode observed by Raman scattering, ^ 3.81 THz (=127
cm~^),0| and that for amorphous GeTe observed by co-
herent phonon spectroscopy. 'l2l| As the pump power den-
sity increases from 0.8 to 12.7 mJ/cm^, the time period of
the Ai mode increases, corresponding to the red-shift of
the phonon frequency from 3.8 to 3.0 THz, as discussed
later.

As the power density of the single pump pulse in-
creases, the amplitude of the Ai mode increases and
saturates for the highest employed fluence, as shown
in the inset of Fig. 1. Experiments were not per-
formed above the fluence of 13 mJ/cm^ because of sam-
ple damage by laser ablation through generation of dense
electron- hole plasma. A similar saturation for the
phonon properties was observed for the optical phonons
in semimetals 15[ and in semiconductors 16j under sim-
ilar conditions of high-density single pump excitation.
The phonon softening observed with the single pump
pulse may be ascribed to the phonon self-energy effect [17 1
or the electronic softening. [16[ The observation of un-
derdamped oscillation with single pump pulse suggests
that the GeTe crystal stays far from the critical point
of the phase transition. In order to drive the Ai mode
closer to the critical point while avoiding saturation
and sample damage, the multiple pump pulse excitation
technique 13, 3] was applied to the Ai mode.

By adding two mirror arms to the Michelson
interferometer, THz-rate pulse train with four pulse
sequence was generated at a variable repetition rate, as
shown in Fig. 2 (a). By moving the mirrors of arms in
the twin interferometer, the separation time Atij (i,j =
1, 2, 3, 4) between the pulse components of Pi and Pj
was controlled. The time derivatives of the transient re-
flectivity changes obtained by using the multiple pump
pulses are shown in Fig. 2(b). Here, the power density
of each pump pulse is 3.8 mJ/cm^, such that the maxi-
mum total pump power (3.8 mJ/cm^ x 4) exceeds that of
the single pump excitation (12.7 mJ/cm^) without sam-
ple damage. We aimed that each pump pulse force the

3

0.0 0.4 0.8 1.2 1.6 2.0
Delay time (ps)

FIG. 2: (a) The optical layout of a twin Michelson interfer-
ometer for the generation of the pulse train. BSs are the
beam splitters. Each mirror arm labeled Pi (i=l, 2, 3, 4)
is computer-controlled, (b) Repetitive excitation of the Aig
mode using THz-rate pulse train. Pi, P2, P3, P4 are first,
second, third, fourth pump pulses, respectively. Ati2, At23,
and At34 are set to be 290 fs, 320 fs, and 345fs, respectively.
The open circles represent the experimental data. The solid
lines represent the fitting of the time-domain data after the
final pump pulse using Eq. (1).

J I I I I I L_

4 8 12

Pump power (mJ/cm )

FIG. 3: The pump power dependence of the frequency of the
Ai mode obtained for the single pump pulse and the pulse
train, respectively. The dotted curve is an eye guide and the
solid line corresponds to a fitting of the data with a linear
function.

oscillation of the coherent Ai mode through the repeti-
tive excitation. In order to well drive the soft phonon to
an in-phase motion, the time intervals in the THz-rate
pulse train was set to be unequal; as shown at the top
of Fig. 2(b), the pulse delay was increased for each sub-
sequent pulse to match the increased pulse intervals for
each driven phonon oscillation. The time-domain data in
Fig. 2(b) clearly demonstrate both an enhancement in
the amplitude of the coherent Ai mode by a repetitive
excitation and a drastic decrease in the dephasing time
of the coherent phonon. The data of Fig. 2(b) was fitted
with Eq. (1), in order to obtain the dephasing time of
the coherent phonon and the frequency of the Ai mode
for various total pump powers.

Figure 3 shows the obtained frequency of the Ai mode
as a function of the total pump power together with the
results for single pump excitation. The Ai frequency de-
creases linearly from 3.8 to 2.5 THz as we increase the
number of pulse sequence, while the single pulse excita-
tion produce saturation at highest fuences. The dephas-
ing time decreases from 570 to 180 fs as the number of
pulse sequence is increased. The largest damping rate of
7 ~ 1.8 ps~^ (=1/(180 fs)/7r) observed with four pump
pulses is significantly close to the frequency of uj ~ 2.5
ps~^ (or 2.5 THz), showing that the Ai mode is driven
to near the point of the critical damping (w = 7).

We note that the lowest frequency of the soft mode ob-
served in the present study using THz-rate pulse train,
of 2.5 THz, is comparable to the incoherent phonon fre-
quency at a temperature of ^ 590 K, according to the
Curie- Weiss law. JtJ This confirms that the crystal lat-
tice is really close to the transition point.

4

In conclusion, wc generated a large amplitude coher-
ent phonon in GeTe ferroelectrics by use of an intense
THz-rate pulse train, whose time period was matched
to the increased phonon period. The soft phonon fre-
quency decreases linearly with the number of the pulses
in the excitation sequence, leading to near critical damp-
ing condition for a 4 pulse excitation. The optical con-
trol of the lattice vibration will initiate a new crystal
structure, which cannot be observed by conventional fre-
quency methods. Such experimental scheme for manipu-
lating collective atomic motions can be generally pursued
in physics, biology, and chemistry.

The authors acknowledge T. Dumitrica and P. R.
Vinod for critical reading of the manuscript. We thank
the Venture Business Laboratory of Osaka University
where the first stage of the piimp-probe experiments were
carried out. The authors are grateful for the support re-
ceived through a Grant-in- Aid for the Scientific Research
from the Ministry of Education, Culture, Sports, Science,
and Technology of Japan.

[1] A. M. Weiner, D. E. Lcaird, G. P. Wiederrecht, and K.

A. Nelson, Science 247, 1317 (1990).
[2] T. Dekorsy, W. A. Kiitt, T. Pfeifer, and H. Kurz, Euro-

phys. Lett. 23, 223 (1993).
[3] M. Hasc, K. Mizoguchi, H. Harima, S. Nakashima, M.

Tani, K. Sakai, and M. Hangyo, Appl. Phys. Lett. 69,

2474 (1996).

[4] U. Ozgiir, C-. W. Lee, and H. O. Everitt, Phys. Rev.

Lett. 86, 5604 (2001)
[5] Y. Liu, S-. G. Park, and A. M. Weiner, Opt. Lett. 21,
1762 (1996).

[6] M. Okuda, H. Naito, and T. Matsushita, Jpn. J. Appl.
Phys. Part 1, 31, 466 (1992).

[7] E. F. Steigmcior and G. Harbeke, Solid State Communi-
cations 8, 1275 (1970).

[8] W. Cochran, Phys. Rev. Lett. 3, 412 (1959).

[9] K. M. Rabc and J. D. Joannopoulos, Phys. Rev. B 36,
6631 (1987).

[10] H. Z. Cummins and A. P. Levanyuk, in Light Scatter-
ing Near Phase Transition (North-Holland, Amsterdam,
1983).

[11] T. P. Dougherty, G. P. Wiederrecht, K. A. Nelson, M. H.

Garrett, H. P. Jensen, and C. Warde, Science 258, 770

(1992); Phys. Rev. B 50, 8996 (1994).
[12] M. Forst, T. Dekorsy, C. Trappe, M. Laurenzis, and H.

Kurz, Appl. Phys. Lett. 77, 1964 (2000).
[13] H. J. Zeiger, J. Vidal, T. K. Cheng, E. P. Ippen, G.

Dresselhaus, and M. S. Dresselhaus, H. J. Zeiger, Phys.

Rev. B 45, 768 (1992).
[14] K. Sokolowski Tinten, J. Bialkowski, A. Cavalleri, D. von

der Linde, A. Oparin, J. Meyer-ter-Vehn, and S. L Anisi-

mov, Phys. Rev. Lett. 81, 224 (1998).
[15] M. F. DeCamp, D. A. Reis, P. H. Bucksbaum, and R.

Merlin, Phys. Rev. B 64, 092301 (2001).
[16] S. Hunsche, K. Wicnecke, T. Dekorsy, and H. Kurz, Phys.

Rev. Lett. 75, 1815 (1995).
[17] M. L. Ledgerwood and H. M. van Driel, Phys. Rev. B

54, 4926 (1996).
[18] G. P. Wiederrecht, T. P. Dougherty, L. Dhar, K. A. Nel-
son, D. E. Leaird, and A. M. Weiner, Phys. Rev. B 51,

916 (1995).

[19] A. M. Weiner, Rev. Sci. Instrum. 71, 1929 (2000).